基于小波包变换与粗糙集的船舶电机定子故障诊断方法

Stator fault diagnosis method for marine electrical motors based on wavelet packets transform and rough set theory

为在少量的传感器以及无先验故障知识的情况下,实现船舶电机定子故障状态的确定以及故障严重程度的划分,提出了对定子电流进行小波包变换提取故障特征,基于粗糙集理论建立可分辨矩阵,并从中提取故障诊断规则的方法.通过对电流特征信号进行小波包分解,使相应分解子频段能始终覆盖随电机转差率变化的故障特征频率;利用子频段重构系数的均方根值变化率作为特征指标,实现故障特征的提取.采用自组织特征映射网络对特征指标进行聚类,由相邻子频段的均方根值变化率组成一组学习样本的方法,减少了子频段之间混叠问题对于聚类结果的影响.实验室条件下进行了电机定子故障的实验,通过对故障数据的应用,验证了该方法的可行性.

To discern fault evolution and criticality for marine motors under less sensors and without a priori failure diagnosis knowledge conditions, the discernibility matrix was established and a extraction method for diagnosis rules was developed based on rough set theory. The stator current was decomposed by wavelet packets. The decomposed sub-frequency bands could always cover the fault eigenfrequencies varying together with the slip. The fault eigenvalues were obtained by the mean-squared root method using reconstructed node coefficients. The eigenvalues were clustered based on a self-organizing feature map. The disturbance owing to wavelet overlap was significantly decreased using training samples composed of two mean-squared roots for the adjacent nodes. Under the laboratory condition, a artificial motor fault experiment was designed. Experimental results validate the feasibility of the proposed method.